visNetwork

Data Preparation

As with other network visualization frameworks, the input must include data organized as a collection of nodes and edges (connections between the nodes). For visNetwork, this data should be tabular (i.e. a data.frame in R).

The minimal example from the visNetwork() help file:

nodes <- data.frame(id = 1:3)
edges <- data.frame(from = c(1,2), to = c(1,3))

nodes

##   id
## 1  1
## 2  2
## 3  3

edges

##   from to
## 1    1  1
## 2    2  3

Note the column names. Those names are strictly required when defining nodes/edges. The edge identifiers (“from” and “to” columns) must contain values that also appear in the identifiers in nodes (“id” column).

Demonstrating the features of the package is difficult with such a minimal dataset. The code below will prepare some simulated data to be visualized. The example will involve nodes with alphanumeric identifiers (e.g. “U89”), edges that represent connection between some (but not all) of the nodes, as well as a categorization of the degree of that connection (here stored in the “feeling” column). The code is further documented inline:

## load packages
library(tidyverse)
library(visNetwork)

## make sure to set seed for reproducibility
set.seed(20200329)

## write a function to create simulated identifiers
gen_id <- function() {
  
  ## create a random alphanumeric combination
  paste0(sample(LETTERS, size = 1),
         sample(10:99, size = 1),
         collapse = "") 
}

## create a vector of identifiers using the function defined above
samples <- sapply(1:20, function(x) gen_id())

## double check that these are all unique
samples <- unique(samples)

## create a tibble with color palette corresponding to a category
pal <- 
  tibble(feeling = c("love","hate","meh"),
         color = c("lightblue","firebrick","goldenrod"))

## create tibble for the nodes
## id column will have sample ids
nodes <- tibble(id = samples)

## create tibble for edges
edges <- 
  ## define all possible combinations of sample ids
  combn(samples, 2) %>%
  ## transpose
  t(.) %>%
  ## coerce the matrix to a tibble
  as_tibble(.) %>%
  ## set the column names
  set_names(c("from","to")) %>%
  ## create new column that qualifies the nature of the connection
  ## here we're calling this "feeling"
  ## if it is NA the connection does not exist
  mutate(feeling = sample(c("love","hate","meh", NA), 
                          size = nrow(.),
                          prob = c(0.2,0.2,0.2,0.5),
                          replace = TRUE
                          )) %>%
  ## if the feeling is NA then the connection does not exist ... filter out
  filter(!is.na(feeling)) %>%
  ## make sure the same ID doesn't appear in from AND to
  filter(from != to) %>%
  ## join in the palette tibble defined above
  left_join(pal)

## take a look at the simulated nodes
head(nodes)

## # A tibble: 6 x 1
##   id   
##   <chr>
## 1 K93  
## 2 P28  
## 3 P55  
## 4 G18  
## 5 V50  
## 6 L49

## take a look at the simulated edges
head(edges)

## # A tibble: 6 x 4
##   from  to    feeling color    
##   <chr> <chr> <chr>   <chr>    
## 1 K93   P28   hate    firebrick
## 2 K93   V50   love    lightblue
## 3 K93   H38   meh     goldenrod
## 4 K93   J69   meh     goldenrod
## 5 K93   D77   love    lightblue
## 6 K93   B67   love    lightblue

Creating the visNetwork plot

With data prepared accordingly, creating the default plot is as simple as calling the visNetwork() function:

visNetwork(nodes,edges)

Labeling nodes and edges

The nodes and edges displayed above can be labelled … so long as the corresponding data frames include “label” columns:

## for this example use the id as label for nodes
nodes <-
  nodes %>%
  mutate(label = samples)

## create a new column that qualifies one of the "feeling" levels
edges <-
  edges %>%
  ## add another column to further qualify one of the "feeling" levels
  mutate(label = ifelse(feeling == "meh", 
                          sample(c("+","-")),
                          NA))

visNetwork(nodes,edges)

Additional options and layout

The initial plot can be further customized by subsequent functions. Using the %>% operator can help with the legibility of the code.

visOptions() has arguments for the height and width of plot, as well as miscellaneous features to configure behavior when clicking and interacting:

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE)

The default layout settings will randomly configure nodes each time the plot is generated. The randomSeed argument to visLayout() allows the coordinates to be fixed:

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
  visLayout(randomSeed = 1999)

The layout of the nodes can be customized beyond the default. For example, the network can be arranged in a hierarchy (like a tree) or in a circular format (the default to visIgraphLayout() is “layout_nicely”):

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
  visLayout(randomSeed = 1999) %>%
  visIgraphLayout()

Including a legend for edges

Either nodes or edges can be annotated and represented in a legend with visLegend(). In this example, the edges will be annotated with the “feeling” column:

edges_legend <- 
  pal %>%
  rename(label = feeling)

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
  visLayout(randomSeed = 1999) %>%
  visIgraphLayout() %>%
  visLegend(addEdges = edges_legend)

While the legend in the plot corresponds to the colors of the edges, there are arrows even though the network is not directed. To fix this the data.frame for the edges legend needs to include a list column with instructions to disable the arrows:

edges_legend <- 
  pal %>%
  rename(label = feeling) %>%
  mutate(arrows = list(to = list(enable = FALSE)))

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
  visLayout(randomSeed = 1999) %>%
  visIgraphLayout() %>%
  visLegend(addEdges = edges_legend)

Using icons for node shapes

visNetwork incorporates icons from Font Awesome for styling nodes. To implement, call the addFontAwesome() function and then modify the shape and icon arguments to visNodes():

visNetwork(nodes, edges) %>%
  visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
  visLayout(randomSeed = 1999) %>%
  visIgraphLayout() %>%
  visLegend(addEdges = edges_legend) %>%
  addFontAwesome(name = "font-awesome") %>%
  visNodes(shape = "icon", icon = list(code = "f007", color = "black"))